Two sequences are defined as overlapping if one is a subsequence of the other. The fraction of ISREs that overlap with a class of hexamers is indicated in filled black bars. The white bars represent the mean fraction of 1,000 comparisons of ISREs with sets of random hexamers of equal size as the class (error bars depict one standard deviation). Refer to Protocol S1 on details of the compilation of known splicing regulatory sequences from multiple sources. “No SS” refers to using ISREs without splice-site like sequences.

The 3′ss reporter assay is depicted on the right. Boxes indicate exons, horizontal lines connecting the exons indicate introns; angled lines connecting boxes represent splicing choices. The nucleotide sequence between the intron-proximal and intron-distal splice sites (indicated by arrows) are depicted. ISREs indicated as a black bar are inserted in tandem duplicates between the competing 3′ss (see legend to Figure 2).

The lengths of the horizontal bars represent the maximum Z score for the ISRE (see Table S3 for all ISREs enriched near AS exons; see Protocol S1 for Z score calculation). The blue bars correspond to the downstream ISREs, represented by PWMs on the far left. PWMs were generated for each ISRE by aligning all instances of sequences that make up the ISRE in the introns proximal to exons (see Protocol S1 for details). The red bars correspond to the upstream ISREs, represented by PWMs on the far right. The PWMs are organized from most enriched (top) to least enriched (bottom) near AS exons. PWMs are used here for visualization purposes.

A gene is defined as differentially expressed in a specific tissue or cell line if its expression in the tissue or cell line is two standard deviations above the median expression across all tissues/cell lines. We denote genes that contain ISRE-proximal exons as ISRE genes. For each ISRE, the tissues or cell lines in which the ISRE genes were significantly enriched (positive value) or depleted (negative value) are illustrated, followed by the enriched or depleted GO terms. The y-axis is the log₁₀ p value for significance. For notational simplicity, a minus sign indicates significantly depleted values. Please refer to Protocol S1 for details on controlling for background conservation in the intronic sequences.
(A) Downstream ISREs.
(B) Upstream ISREs.

The 5′ss reporter assay is depicted on the right. Boxes indicate exons, horizontal lines connecting the exons indicate introns; angled lines connecting boxes represent splicing choices. The nucleotide sequence between the intron-proximal and intron-distal splice sites (indicated by arrows) is depicted. ISREs indicated as a black bar are inserted in tandem duplicates between the competing 5′ss. Each ISRE-containing reporter was transfected in triplicate in human 293T cells. A representative PCR product for each ISRE-containing reporter is depicted on the etimidium bromide–stained agarose gel. The larger and shorter bands on the gel correspond to the intron-proximal and intron-distal isoforms, respectively (indicated by arrows on the right of the gel). Horizontal bars depict the mean of three replicates measuring the ratio between the intron-distal isoform to the sum of both isoforms. Error bars depict one standard deviation. ctl indicates individual control elements. Asterisks on the right of each bar indicate ISREs that significantly altered splicing relative to the averaged controls by a t test (p < 0.01). Please refer to Protocol S1 for details on quantification of the isoforms.

Conserved positional bias plots of ISREs in the intronic regions upstream (black line) and downstream (red dashed line) of all exons (see Materials and Methods). The PWM representation of the ISREs is depicted above their respective plots.
(A) ISRE U25 and D20 resembled the FOX1 binding site (TGCATG) and showed significant conserved enrichment throughout the 400 bases flanking exons, as well as prominent peaks in the region −130 to −40 and +15 to +120 of the intron.
(B) ISRE U22 and D29 resembled the branch signal (TACTAAC). U22 showed a prominent peak at −40 to −10. D29 was significant from +15 to +80.
(C) ISRE U9 and D9 resembled the “CUG” binding site of MBNL, and showed significant conserved enrichment throughout the 400 bases flanking exons. U9 peaked at −60 to 0 and D9 peaked at +0 to +120.

Intronic regions upstream and downstream of exons predicted to be included/skipped at different significance levels (see Protocol S1 for details on the dataset). The significance of enrichment (red boxes)/depletion (green boxes) are reported as log₁₀ p values; a plus sign indicates enrichment, and a minus sign indicates depletion. For example, an ISRE represented by GCATG was enriched in intronic regions downstream of brain-included exons at p < 1 × 10⁻⁴ (reported as 4). The motifs highlighted in yellow have been experimentally verified to alter splice site choice in our reporter constructs.
(A) Downstream ISREs enriched/depleted in the downstream intronic regions.
(B) Upstream ISREs enriched/depleted in the upstream intronic regions.

A total of 400 bases of intronic region flanking each internal exon were extracted for each splicing factor. Each box in the grid represents the total number of nonoverlapping occurrences of the conserved ISRE (either upstream ISRE or downstream ISRE) in the flanking intronic regions. ISREs that occur at least once in the flanks of at least one exon are retained in the grid. The exon–intron structure of selected exons are depicted below each grid. Tall vertical rectangles depict exons, and interconnecting lines represent introns; the line connecting nonadjacent exons represents alternative splicing of the middle exon. Red bars proximal to the alternative exon illustrates the ISRE. The level of conservation across multiple species (from the hg17 UCSC genome browser) is depicted below the exon.
(A) Nova1: U149 (TTTCAT) and D150 (TTTCAT) are the most abundant ISREs in the flanking intronic region of exon 4. The flanking intronic region of exon 4 has the highest sequence conservation across mammalian genomes among all flanking intronic regions in Nova1.
(B) FOX1: U25 (TGCATG) and D20 (TGCATG) are the most abundant ISREs in the flanking intronic region of exon 10 in FOX1. The flanking intronic region of FOX1 has the highest sequence conservation among all flanking intron regions in FOX1. The exon 10 sequence (boxed) and 15 bases of flanking intronic regions are depicted below the exon–intron structure of exons 9, 10, and 11. TGCATG sites (in red) are overlapping the 5′ss and 3′ss.
(C) nPTB: exon 10 of nPTB has transcript evidence for alternative splicing, and contained highly abundant U26 (TTCTT), U121 (TGTCT), U122 (TTCCTT), U149 (TTTCAT), D21 (TTCCTT), D108 (TCCTTT), D115 (TGTCT) and D150 (TTTCAT).
(D) hnRNP C: the flanking intronic regions of exon 2 of hnRNP C contains 17 copies of TTTTT (an element in U4), and is the most highly conserved among the intronic regions in hnRNP C.